AI competency is a leadership standard now. Let’s treat it like one.

There is a particular kind of meeting that most people will recognise. Someone senior- experienced, respected, genuinely good at their job in most respects- says some version of the following: “I’ll be honest, I'm just not a technology person.” They say it with a rueful smile and a cheeky shrug, as though it's an endearing personality quirk rather than a professional gap. The room chuckles and nods sympathetically, and the conversation moves on.

For a long time, that was acceptable. ‘Technology comes and goes’. Some of it mattered, some of it didn't, and a leader with strong instincts and good judgment could afford to let the enthusiasts handle the tools while they focused on the work.

I hate to break it to you, but that period is over.

AI is not a new tool in the way that a new CRM or a new project management platform is a new tool. It is a structural shift in how knowledge work gets done- in how decisions get made, how teams operate, how organisations learn and communicate and build. McKinsey's 2024 research found that AI has already affected knowledge worker productivity across every sector they surveyed, and the World Economic Forum projects that AI and automation will reshape more than 40% of working tasks within the next five years. A leader who cannot engage with it at a basic level is not a technology-sceptic. They are operating with a meaningful blind spot in one of the most consequential areas of their job. At this point, the industry's patience with that blind spot has run out, or it should have.

Before we talk about standards and enforcement- and we will- it's worth being precise about what a minimum acceptable AI competency for leaders actually looks like, because most organisations haven't defined it. They've made vague noises about 'AI literacy' and pointed people at a few optional modules, which is roughly as useful as telling someone to 'get more comfortable with finance' and leaving a spreadsheet on their desk.

A minimum standard is not about technical fluency; leaders do not need to understand how large language models work, and they do not need to be prompt engineers. What they need is functional competency: enough understanding to use AI tools as a meaningful part of their daily work, enough critical awareness to evaluate AI outputs rather than accepting or rejecting them on instinct, and enough strategic literacy to understand what AI can and cannot do for their team.

In practice, that means a leader who has reached the minimum standard should be able to do four things.

1. They can use AI to assist with at least some of the cognitive work they do regularly- drafting, summarising, planning, generating options.

2. They can evaluate what comes back critically, understanding that AI output requires judgement rather than trust.

3. They can have an informed conversation with their team about where AI is and isn't appropriate in their domain.

4. And they can model a relationship with the technology that is neither evangelical nor dismissive- curious, practical, and honest about uncertainty.

That is not a high bar. It is a reasonable professional expectation in 2026, and the fact that so many organisations haven't articulated it clearly is its own problem.

Most leaders who are behind on AI competency are not behind because they are incapable. They are behind because nobody has been honest with them about what's expected, and the path to getting there has been unclear, optional, and easy to defer.

The first thing that changes that is making the expectation explicit. Not in a town hall about ‘our exciting AI journey,’ not in a strategy document that circulates once and is never mentioned again- in a direct conversation between a leader and their manager or HR partner, in which it is clearly stated: this is what competency looks like, this is how we'll support you to get there, and this is the timeline. That conversation has to happen, and it has to be specific. Vagueness is how accountability disappears. Research by Gollwitzer on implementation intentions- the behavioural science term for 'if-then' goal-setting- consistently shows that people who commit to a specific plan for when and how they will do something are significantly more likely to follow through than those who simply intend to. The same logic applies here: a vague commitment to 'develop AI skills' will be deferred indefinitely; a specific commitment to use AI for meeting preparation every Monday for the next four weeks will not.

The learning itself needs to be designed for the same people who roll their eyes at mandatory training, because those are exactly the people who need it. That means it cannot be a course. It cannot be a module. It needs to be experiential, low-stakes, and immediately relevant to the work they already do.

The most effective approach I've seen starts with replacing something- not adding something. Find one thing the leader already does regularly and tediously, and show them how AI handles it in five minutes. Not a demo. Them doing it, with support, on a real piece of their own work. The experience of 'that genuinely saved me forty minutes' bypasses months of theoretical persuasion. This is consistent with self-determination theory- Deci and Ryan's foundational research on motivation- which shows that competence, the direct experience of being capable at something, is one of the three core conditions for intrinsic motivation. You cannot argue someone into feeling competent. They have to experience it. From that point, the conversation shifts from convincing them to engage to helping them develop the critical awareness to engage well.

From there, the development needs to address what actually holds experienced leaders back, which is usually not technical anxiety but epistemic discomfort. Leaders who are good at their jobs have spent years developing reliable instincts, and AI asks them to engage with a tool whose outputs they can't fully evaluate yet. That's genuinely uncomfortable. It connects to what psychologists call expert identity threat- the discomfort high performers experience when asked to operate as novices in a domain adjacent to their expertise. Researcher Matthew Yeager's work on identity and learning suggests that framing matters enormously here: people protect their expert identity when they feel it's under attack, and disengage; they engage when the new capability is framed as an extension of existing judgment rather than a replacement for it. Which means it helps to name this explicitly: the skill you're developing is not trust in AI, it's judgment about AI. Those are not the same thing, and the second is a skill you already have the foundations for. Experienced leaders evaluate uncertain information all the time. This is that.

Peer learning accelerates this significantly. A leader who is sceptical of AI will sit through a training session and wait for it to end. The same leader watching a respected peer describe- specifically, without evangelical fervour- how they use it to prepare for difficult conversations, to stress-test a business case, or to get a first draft of a communication done in twenty minutes, will lean forward slightly. Bandura's social learning theory has been demonstrating this since the 1970s: we learn most powerfully through observation of credible models, and credibility is determined by similarity and status, not by expertise alone. A peer who is also figuring this out is a more compelling model than a consultant who has been living and breathing AI for three years. Social proof from people at the same level, doing the same kind of work, matters more than content delivered from the front.

At a certain point, the question of AI competency stops being a development conversation and becomes a performance conversation. That point is when the support has been offered, the expectation has been clearly communicated, the timeline has elapsed, and the gap persists. And the honest truth is that if AI competency is genuinely a leadership requirement- and it is- then not meeting it should have consequences.

This is not about punishing people for being slow to adapt. Most organisations have created the very conditions that make slow adaptation inevitable: unclear expectations, optional learning, no accountability, and a cultural norm that treats AI resistance as a personality trait rather than a professional gap. When those conditions exist, leaders learn quickly that 'I'm not really a technology person' is a complete sentence, and nobody will push back on it.

Changing that requires changing the norm, and norms change through visible consequences. Cialdini's research on social proof and norm behaviour is useful here: people look to what others around them are doing and what happens when they don't comply. If senior leaders who avoid AI development face no discernible consequence, the norm becomes avoidance. If they do- if AI competency appears in performance conversations, informs progression decisions, shapes who gets stretch assignments- the norm shifts. What that looks like in practice will vary: it might mean AI competency becoming an explicit criterion in performance reviews, or a prerequisite for progression conversations, or a factor in how sponsorship opportunities are allocated. It doesn't have to be punitive. It does have to be real. If there is genuinely no consequence for not developing this capability, then it isn't a standard- it's a preference. And preferences are easy to ignore.

The organisations that will get this right are the ones that are honest about two things simultaneously: that they will invest seriously in helping every leader reach the standard, and that reaching the standard is not optional. Both halves of that commitment matter. The first without the second produces good intentions and no movement. The second without the first produces resentment and quiet attrition.

There is a version of this conversation that treats AI competency as something we are imposing on leaders, a new burden to manage on top of everything else- and I want to push back on that framing. Leaders have always been expected to develop their capability in the areas that matter most to their role. When organisations moved to data-driven decision making, leaders who couldn't read a dashboard became a liability. When remote work became structural, leaders who couldn't manage distributed teams became a problem. The expectation evolved because the work evolved, and this is just the same.

The leaders who will be most effective over the next decade are not the ones who can explain transformer architecture. They are the ones who have developed genuine judgment about where AI makes their thinking better, where it introduces risk, and what it means for the people they lead. That judgment comes from use- from regular, purposeful engagement with the technology in the context of real work, over time.

Getting leaders to that point is entirely achievable. Letting them opt out of it is no longer a kindness, but a disservice- to them, to their teams, and to organisations that cannot afford to have senior decision-makers operating with a structural gap in one of the most important areas of contemporary leadership.

The bar is not high and it is clear. It is time to say so plainly, support people to meet it, and take seriously what happens when they don't.

Alice Veitch is a Learning & Development Lead with a background in behavioural science. She has led learning strategy across EMEIA and is a respected disruptor in the field.